SVM/ Dual SVM math derivation, non-linear SVM, kernel function详细

最新推荐文章于 2020-05-01 17:37:20 发布

Jay_Tang

最新推荐文章于 2020-05-01 17:37:20 发布

阅读量1.3k

点赞数 7

分类专栏：机器学习核心推导文章标签：数据挖掘机器学习算法

本文链接：https://blog.csdn.net/Jay_Tang/article/details/105154125

版权

本文深入探讨支持向量机（SVM），从线性SVM的概念、目标优化问题到软间隔与 hinge 损失函数，接着介绍非线性SVM如何通过核函数解决高维问题，如多项式、高斯和径向基函数（RBF）等常见核。最后，阐述SVM的双问题和Mercer定理的重要性。

摘要由CSDN通过智能技术生成

往期文章链接目录

文章目录

Linear SVM

Idea

We want to find a hyper-plane $w^\top x + b = 0$ that maximizes the margin.

Set up

We first show that the vector $w$ is orthogonal to this hyper-plane. Let $x_1$ , $x_2$ be any element on the hyper-plane. So we have $w^\top x_1 + b = 0$ and $w^\top x_2 + b = 0$ . Then $w^\top (x_1 - x_2) = 0$ , which implies $w$ is orthogonal to the hyper-plane. Now, we set two dashed lines to $w^\top x + b = 1$ and $w^\top x + b = -1$ . In fact, “ $1$ ” doesn’t matter and we can pick any value here. “ $1$ ” is just the convention.

Now we pick any line parallel (orthogonal) to $w$ (hyper-plane), then the line intersect two dashed line with point $x^{(+)}$ and $x^{(-)}$ . We want to maximize the margin